This project is a collaboration between two neurobiologists and two applied mathematicians. Its main goal is to combine experiment and computation to develop a detailed understanding of how the biophysical properties of individual retinal neurons and synapses shape the parallel processing of visual information during rod vision. We will characterize experimentally and model computationally the components composing the neural circuitry of the mammalian retina that subserves night (scotopic) vision. The anatomy of this circuitry is well-described: photons absorbed by rod photoreceptors generate neural signals that are distributed to multiple types of retinal ganglion cells (GCs), the output cells of the retina, via a series of interneurons. These interneurons are coupled to each other by both chemical and electrical synapses. Each type of GC has a unique response to light, which is presumed to reflect the properties of the circuitry presynaptic to it. This allows each GC type to encode a different feature of the visual scene, thereby facilitating further abstractions by the higher brain areas that ultimately guide behavior. By characterizing and carefully modeling each component, by assembling the components into a comprehensive model, and by validating and refining the overall model experimentally, we will determine how diverse GC outputs emerge from the biophysical properties of parallel retinal microcircuits. Intellectual merit: The retina is one of the few neural circuits for which output evoked by physiological stimuli is similar in vitro and in vivo, making it amenable to experimental analysis. The sheer volume of information encoded by the retina, however, makes understanding its functional circuitry difficult. In an iterative process of simulation and experiment, we will develop detailed biophysical models of the retinal network utilized for rod vision. In particular: 1) Data to constrain ion channel models will be acquired by direct recording from neurons of interest;2) Chemical and electrical synapses will be characterized directly by recording from pairs of synaptically coupled neurons;data will be used to construct a network model of the rod pathway;3) Compartmental models of different neurons will be constructed using anatomical and physiological parameters determined by experiment;4) The overall network model will be validated and refined using light-evoked recordings from different classes of GCs. The computational model will provide a framework in which experiments probing parallel processing of rod signals within the retinal network may be understood. Moreover, we will explore the parameter space within which the model network operates to develop new and experimentally testable hypotheses about retinal function. Broader Impacts: Validated computational models will be made available via the ModelDB database, a public repository of neuronal models, thereby providing a tool to allow other researchers to develop new experimentally testable hypotheses about retinal processing. More generally, the data obtained through this project and the accompanying computational model will provide a validated example of information processing in neural circuits;such information can serve as an example of possible information processing in other neural systems. The proposed project also will provide neuroscientists and applied mathematicians the opportunity for detailed cross-disciplinary training. Applied mathematics students will acquire a solid background in neuroscience, and neuroscience students will acquire a foundation in quantitative modeling techniques. Teaching materials developed as part of this interaction will be made available to the larger neuroscience community. Additionally, the PIs will pursue opportunities to organize a meeting or symposium (e.g., at the annual Society for Neuroscience meeting) to bring together researchers working on combined experimental and computational studies of neural circuits.